khoiucd / LearningToCompare-Tensorflow

Tensorflow implementation for paper: Learning to Compare: Relation Network for Few-Shot Learning.
MIT License
19 stars 4 forks source link
few-shot-learning meta-learning metric-learning

Relation Network - Tensorflow

Unofficial Tensorflow implementation for Relation Network

Paper: Learning to Compare: Relation Network for Few-Shot Learning.

For official pytorch implementation: LearningToCompare_FSL.

Requirements

Python 3

Tensorflow 2.1.0

Data

For mini-Imagenet, please refer to here.

Pretrained models

All models are trained with 50000 episodes.

Experiment Accuracy
miniImageNet 1-shot 50.06 ± 0.12%
miniImageNet 5-shot 65.34 ± 0.86%

Reference

LearningToCompare_FSL.